减小运行时优化开销的方法  

Ways to Reduce the Cost of Runtime Optimization

作　　者：郭振宇[1] 刘利[1] 陈彧[1] 汤志忠[1] 

机构地区：[1]清华大学计算机科学与技术系

出　　处：《计算机工程》2006年第24期63-65,120,共4页Computer Engineering

摘　　要：运行时优化在程序运行期间,根据采集到的相关信息,确定程序的热点并进行优化,从而加速程序的执行。然而,运行时优化本身有一定的开销,有时候会抵消甚至超出优化得到的效果。该文设计和实现了一个基于SMP/IPF(英特尔安腾系列)/Linux架构的自适应二进制代码优化/编译框架,其中包含了运行时优化。分析了运行时优化的阶段和开销,并介绍了在设计和实现该框架的过程中,为减小这种开销所提出的思路和采用的方法。Runtime optimization analyzes the runtime intbrmation it collects, identifies hot spots, applies optimization on them, thus speedups the execution of the programs. However, the system itself may consume critical resources, which sometimes counteracts or even outweighs the benefit it gains, and leads to the failure of the optimization. This paper implements an adaptive binary optimization/compilation framework on SMP/IPF （Jntel Itanium processor family）/Linux, among which runtime optimization is included. It also analyzes the stage and cost of runtime optimization, introduces the ideas and ways which are conducted to reduce them, during the implementation of the framework.

关 键 词：运行时优化 优化开销 持续优化 

分 类 号：TP314[自动化与计算机技术—计算机软件与理论]